You are on page 1of 5

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056

Volume: 04 Issue: 02 | Feb -2017 www.irjet.net p-ISSN: 2395-0072

Prevention of Spoofing Attacks in Face Recognition System Using


Liveness Detection
Kewal Bhat1,Suryapratap Chauhan2,Gopal Benure3,Prafulla Ambekar4,Prof. Sagar Salunke5
1,2,3,4 BE Students, Dept. of Computer Engineering, PCCOE, Pune, Maharashtra, India
5Professor, Dept. of Computer Engineering, PCCOE, Pune, Maharashtra, India

---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Biometric system has gained wide selection of 1.INTRODUCTION
motivations and applications in security domain. Biometric
systems relay on the biometric characteristics/data taken
In this tightly connected networked society, personal
from the user for authentication. sadly such biometric identification has become critically necessary.
information is stolen or duplicated by the Biometric identifiers are commutation ancient
imposters/unauthorized users. Most of the biometry systems identifiers, because it is troublesome to steal, replace,
rely strictly on distinguishing the physiological characteristics forget or transfer them. A 2D-image primarily based
of the user. It becomes easier to spoof in these biometric facial recognition system will be simply spoofed with
systems with the help of faux biometric it any reduces the straightforward tricks and a few poorly-designed
dependability and security of biometric system. Spoof fools the systems have even been shown to be fooled by the
system through the method of deception and impersonating
imposters. Spoofing with photograph or video is one
others to create out that they're licensed so as to achieve
among the foremost common manners to circumvent a
access in to the biometric system. Now a days spoofing has
become quite common on the net that therefore ends up in
face recognition system.
determine stealing and fraud. There are several level of
Liveness detection mistreatment facial expression in
spoofing attacks like putting faux biometry on the detector,
biometric system may be a technique to capture the
replay attack, attacking the entrance centre corrupting the
intermediator, attacking the application etc. These image of the person and take a look at for his/her
successively can cut back the extent of security and aliveness when obtaining documented. Automatic
dependability of biometric system. liveness identification using extraction of caput and face boundaries and facial
the facial expression also has been receiving a lot of attention features is vital within the areas of face recognition,
compared to other biometric modalities. prevention of spoof criminal identification, security and police
attack in biometric system is done by detecting the liveness of investigation systems, human pc interface, and model-
the user with the assistance of native facial expression like eye based video writing. In general, the processed face
blinking, lip movement, forehead and chin movement pattern recognition includes four steps. First, the face image is
of the face detected with real-time generic web-camera. within
increased and segmental. Second, the face boundary
the planned work, a good authentication system using face
and facial expression square measure detected. Third,
biometric modality by developing the aliveness detection
model using the variations within the facial movements. the extracted options square measure matched against
the options within the information. Fourth, the
Key Words: Biometrics, Face Recognition, Liveness classification or reorganization of the user is achieved.
detection, Template matching, Spoof attack Further, aliveness of the user is to be tested in-order to
forestall the spoof attack. Providing dependableness
associate degreed security within the biometric system
has become a need of an hour. Since the present
biometric systems designed mistreatment many
strategies and algorithms fails to beat the fraud and
larceny identity. It becomes necessary to make an

2017, IRJET | Impact Factor value: 5.181 | ISO 9001:2008 Certified Journal | Page 554
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056
Volume: 04 Issue: 02 | Feb -2017 www.irjet.net p-ISSN: 2395-0072

extremely secure and reliable biometric system that is photograph or video is one among the foremost
spoof free employs the facial characteristics variations common manners to attack a face recognition system.
of person to beat spoof attack by detecting the Automatic facial feature extraction, is one of the
aliveness. It uses the aliveness detection method that foremost necessary and tried issues in computer
successively uses strategies like Viola Jones for face vision. Aliveness discovering is that the ability to detect
detection, LBP for feature extraction and Manhattan artificial objects given to a biometric device with
Distance classifier to spot the genuineness of the user associate degree intention to subvert the recognition
and variations within the facial expression to prevent system. The paper presents the information of iris
spoof attack. User authentication is that the basic output signal pictures with a controlled quality, and its
demand of any security system. Facial biometrics is basic application, specifically development of aliveness
very difficult biometric modality as face is acquired detection technique for iris recognition. Single image-
remotely. The aliveness detection using facial based face aliveness detection technique for
movements to stop the spoof attack is also rising discriminating 2-D masks from the live faces. Still
technique. Most of the researchers are creating their pictures taken from live faces and 2-D masks were
efforts to style such systems. The literatures available found in reality the variations in terms of form and
for these are summarized below. The identity spoofing detailed. Face Liveness detection from one Image with
may be a competitor for high-security face recognition thin low rank additive discriminative model. Spoofing
applications. With the appearance of social media and with photograph or video is common technique to
globalized search, face pictures and videos are wide- avoid a face recognition system. A real-time and non-
spread on the web and might be probably used to intrusive method to handle face aliveness relies on
attack biometric systems while not previous user individual pictures from a generic web camera. A real-
consent . biometric authentication system for time Liveness detection approach against photograph
mechanically identifying or verifying an individual spoofing in face recognition, by recognizing
from a digital image or a video frame from a video spontaneous eye blinks, that may be a non-intrusive
supply. Euclidian distance take a look at is used for manner. The approach needs no additional hardware
checking a persons aliveness that ensures the aside from a generic net camera. Eye blink sequences
detection of fake/dummy pictures. Face recognition usually have a fancy underlying structure.
systems don't seem to be able to work with arbitrary
input pictures taken below completely different 1.1 Proposed System:
imaging conditions or showing occlusions and/or The proposed model provides the protection to
variations in expression or cause. To support face biometric system by authenticating the user with face
recognition one must perform a face image alignment attribute along with aliveness detection using
(normalization) step that takes occlusions/variations variations in facial eye, lip, chin and forehead
into consideration. The face detection technique is movements. The designed model is reinforced by
based on coloring data and fuzzy classification. a providing the protection in 2 phases i.e. performing
replacement algorithmic rule is projected so as to authentication and aliveness checks. The designed
discover automatically face options (eyes, mouth and system for the projected system is as shown within the
nose) and extract their correspondent geometrical Fig.1. Commencement within the planned model is
points. It is exploited for motion analysis onsite to getting the image of face biometric modality. Further,
verify Liveness likewise on accomplish lip reading of localization of facial portion is to be carried using Viola
digits. A methodological novelty is that the Jones methodology. The feature extraction is that the
recommended quantized angle options being designed vital steps in any biometric system. Extract the native
for illumination invariance without the requirement for regions of the detected face and find eyes, lips,
preprocessing (e.g., bar chart equalization). There's ton forehead and chin locations to extract the options using
of security threat because of spoofing. Spoofing with native Binary Pattern (LBP) operator. The extracted

2017, IRJET | Impact Factor value: 5.181 | ISO 9001:2008 Certified Journal | Page 555
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056
Volume: 04 Issue: 02 | Feb -2017 www.irjet.net p-ISSN: 2395-0072

feature vectors i.e. template are to be hold on firmly


within the information. Hence, construct the templates
from extracted options individually. Throughout
identification, compare the stored template from the
information with the generated feature vector of the
user using template matching i.e. with Manhattan
distance. If matching is successful then perform the
aliveness check using the variations in native regions of
facial features like eyes, lips, forehead and chin. If
there's a variation in these native features, then user is
alive else user isn't alive.

B. Face Detection and Alignment

The basic need for face recognition is face detection.


Face detection takes place because the camera detects
the image of the user. System object is formed to
observe the placement of a face in an input face image.
The cascade object detector uses the Viola-Jones
detection algorithmic program for face detection. By
default, the detector is designed to detect faces. using
cascade object face region is tracked and with the
assistance of extra properties like bounding box,
tracked face region is delimited with parallelogram
box. Face Detection and tracking is shown in figure.3.
Face alignment is shown in figure 4.

Fig.3.Face Location and tracking

Fig.4.Face Alignment
A. Image Acquisition
C. Feature Extraction
Acquire the facial image of the user using the web camera.
This section is especially required since it acts as an input for The features area unit extracted using LBP methodology
the registration section. Sample face pictures registered wherever every facial image i.e. 256x256 component
within the database is shown in the below figure. resolutions is split into 256 cells (16x16 rows and columns
respectively). The LBP operate is applied to every block of
the face image. The feature vector is made from all the 256
2017, IRJET | Impact Factor value: 5.181 | ISO 9001:2008 Certified Journal | Page 556
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056
Volume: 04 Issue: 02 | Feb -2017 www.irjet.net p-ISSN: 2395-0072

grey values computed from the bar chart generated by the Step 5: Normalize the bar chart.
individual instances of the face pictures. From each user six
instances of face pictures area unit used for coaching. Hence, Step 6: Concatenate normalized histograms of all cells to get
the dimensions of model for one hundred users is 600x256. the feature vector for the window.
The bar chart of the face image is shown in Figure 5.
Algorithm for aliveness Detection:

Step1: Acquire the input as face image and localize the face
i.e. observe face

Step2: find the facial centre by putting a kind of marker

Step3: Draw the virtual line on the centre

Step4: find native eye region of face using equations (Refer


Once face detection, alignment and options extraction are
figure half dozen.)
done successfully, the authentication is used with matching
the users facial feature vector with the model from the i) Estart = ceil(x/2-(x-0.8*x) wherever the worth of x=100
stored database. because the user identification is one-to-
many matching, the feature vector of the individual are ii) end = ceil(x/2+5)
compared with the feature vectors of each individual hold on
within the model database with Manhattan distance. Finally Step 5: find native lip region of face using equations (Refer
the simplest Matching facial image is known using the figure seven.)
minimum Manhattan distance. The Manhattan distance is
i) Lstart = ceil(x-x/4) wherever x=100
computed using following equation
ii) Lend = ceil(x-20)
n
d =| xi - yi| Step 6: find native forehead region of face using equations
i=1 (Refer figure eight.)
Where n=256 is that the dimension of the feature vector, xi is
the i-th element of the sample feature vector, and yi is the I i) Fstart = Estart
the element of the model feature vector. more the aliveness
check is carried for genuine user. If the users authentication ii) Fend = twenty
is failing, no aliveness check is performed.
Step 7: find native chin region of face using equations (Refer
C. Liveness Detection figure nine.)

Local Binary Pattern (LBP) algorithmic program: i) Cstart=Lend

Step 1: Divide the aligned face image into 256 cells. (e.g. ii) Cend=ceil(x)
16x16 pixels for every cell).
Step 8: Convert every RGB native facial feature into grey
Step 2: for every component in a very cell is compared with image.
its eight neighbouring pixels on a circle in clockwise or
counter clockwise direction.

Step 3: If the middle pixel's value is larger than its neighbour, Step 9: Set the threshold worth for every native facial
then label with "1" otherwise, label with "0" i.e. 8-digit feature.
binary range (converted to decimal for convenience).
Step 10: Extract the perimeters of every native facial feature.
Step 4: reason the bar chart, over the cell, of the frequency of
Step 11: notice the mean and variance of every feature using
every "number" occurring (i.e., every combination of that
below equations
pixels area unit smaller and that area unit larger than the
center). Mean = (X) / N

2017, IRJET | Impact Factor value: 5.181 | ISO 9001:2008 Certified Journal | Page 557
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056
Volume: 04 Issue: 02 | Feb -2017 www.irjet.net p-ISSN: 2395-0072

Where, REFERENCES

X = Individual knowledge points [1] Liveness Detection in Biometrics By Maximilian Krieg


and Nils Rogmann.
N = Sample size (number of information points) [2] A Leap Password based Verification System Aman
Chahar *, Shivangi Yadav *, Ishan Nigam, Richa Singh,
Standard Deviation= X!X Mayank Vatsa IIIT-Delhi, New Delhi.
n!1 [3] An Embedded Fingerprint Authentication System, Ms.
Archana S. Shinde, Prof. Varsha Bendre, Dept. of E&TC,
Pimpri Chinchwad College of Engineering Pune, India.
Where, n is that the range of parts within the sample
X could be a vector X could be a mean of vector. [4] The Leap Motion controller: A view on sign language.
[5] Prevention of Spoof Attack in Biometric System Using
Step 12: If there's a variation in native facial features i.e. Liveness Detection By Sanjeevankumar M. Hatture,
Nalinakshi B. G, Rashmi P. Karchi.
eyes, lips, forehead and chin then the person is alive else
not alive. [6] Corneal Topography: An Emerging Biometric System for
Person Authentication By Nassima Kihal, Arnaud
Polette, Salim Chitroub, Isabelle Brunette and Jean
Step 13: Stop Meunier.
[7] A Basic Design for Adaptive Corneal Topography H.J.W.
Spoelder, F.M. Vos2, D.M. Germans 1 .Division of
Physics and Astronomy Faculty of Sciences,Vrije
Universiteit, De Boelelaan
[8] [8] Liveness Detection Technique for Prevention of
Spoof Attack in Face Recognition System Nalinakshi
B. G1, Sanjeevakumar M. Hatture2, Manjunath
Fig.6.Localization of eye and its variation S.Gabasavalgi3, Rashmi P. Karchi4.
[9] Automated Attendance Management System
using Face Recognition.

Fig.7.Localization of lips and its variation

Fig.8.Localization of forehead and its variation

Fig.9.Localization of chin and its variation

3. CONCLUSION

Hence, by introducing liveness detection in face recognition


system, we will be able to overcome the drawbacks of
conventional face recognition system and provide access to
legitimate users only.

2017, IRJET | Impact Factor value: 5.181 | ISO 9001:2008 Certified Journal | Page 558

You might also like